Summary of the invention
For overcome the deficiencies in the prior art, one of the objects of the present invention is to provide a kind of potential customers' recognition methods,It can the data information daily to user be analyzed and processed, obtain user personality label between product information that user is purchasedMatching relationship, thus by artificial intelligence identification model carry out potential customers intelligent recognition, help product dealer realizePrecision marketing is avoided being wasted in the time with some invalid clients, be improved efficiency, save the cost, to improve effectiveness of operation.
The second object of the present invention is to provide a kind of electronic equipment, can the data information daily to user analyzeProcessing, the matching relationship for obtaining user personality label between product information that user is purchased, to pass through artificial intelligence identificationModel carries out the intelligent recognition of potential customers, helps product dealer to realize precision marketing, avoids for the time being wasted in some nothingsIt imitates with client, improves efficiency, save the cost, to improve effectiveness of operation.
The third object of the present invention is to provide a kind of computer readable storage medium, the program operation in the storage mediumWhen can the data information daily to user be analyzed and processed, obtain user personality label between product information that user is purchasedMatching relationship, thus by artificial intelligence identification model carry out potential customers intelligent recognition, help product dealer realizePrecision marketing is avoided being wasted in the time with some invalid clients, be improved efficiency, save the cost, to improve effectiveness of operation.
An object of the present invention adopts the following technical scheme that realization:
A kind of potential customers' recognition methods, comprising:
Training sample obtaining step:
Product sales data is obtained, the product sales data includes the information of the product of client's purchase and of clientPeople's information；
The data information of client is obtained from data information platform according to the personal information of client；The data information includes:Interest big data, track big data and/or general lifestyle big data；The interest big data includes that number is browsed on the line of userAccording to the track big data includes the consumption data of user, and the general lifestyle big data includes the living habit number of userAccording to；
Corresponding label data is established according to the data information of client, the label data includes multiple user personality marksLabel；
The information for the product that all user personality labels of each client and the client buy is matched and as oneA training sample is stored in training set；
Model construction and training step:
According to user personality label, artificial intelligence identification model is established using machine learning algorithm；
The artificial intelligence identification model is trained so that the artificial intelligence using the training sample in training setIdentification model can judge whether the user is potential customers according to the user personality label inputted；
Further, further include identification step:
The data information that user to be identified is obtained from data information platform, according to the foundation pair of the data information of user to be identifiedThe label data answered, and by label data input in trained artificial intelligence identification model with judge the user whether bePotential customers.
Further, described according to user personality label, artificial intelligence identification model tool is established using machine learning algorithmBody are as follows:
Based on XGBoost algorithm, using each user personality label as segmentation attribute, several decision trees of building are integrated intoOne artificial intelligent recognition model.
Further, the pattern function of the artificial intelligence identification model are as follows:
Wherein, t is the quantity of decision tree.
Further, the information of the product includes the brand and model of product；The artificial intelligence identification model energyIt is enough to judge whether the user is potential customers and the Brang Preference of the user according to the user personality label inputted.
Further, further comprise the steps of: when recognizing user is potential customers, according to the personal information of the user andThe data information acquired in the data information platform determines the normal residence of the user for the product dealer couple of corresponding regionThe user carries out precision marketing.
Further, further include negative sample training step:
After training the artificial intelligence identification model, new product sales data is obtained, the product sells numberAccording to the information for the product for including client's purchase and the personal information of client；
The data information of client is obtained from data information platform according to the personal information of client；
Corresponding label data is established according to the data information of client, and label data input is trained artificialRecognition result is obtained in intelligent recognition model；
The recognition result is compared with the effective sale information in the product sales data；
If comparison result is inconsistent, the artificial intelligence identification model is carried out according to the user personality label of the clientAmendment is to improve identification precision.
Further, the user personality label includes: interest tags, zone of action label, level of consumption label, lifePhase tag, personality label and shopping genre labels.
The second object of the present invention adopts the following technical scheme that realization:
A kind of electronic equipment, including processor and memory, the memory is stored with executable computer program, describedProcessor can be read the computer program in the memory and run to realize potential customers' recognition methods as described above.
The third object of the present invention adopts the following technical scheme that realization:
A kind of computer readable storage medium, the computer-readable recording medium storage have executable computer program,The computer program can realize potential customers' recognition methods as described above when running.
Compared with prior art, the beneficial effects of the present invention are:
Potential customers' recognition methods can the data information daily to user be analyzed and processed, obtain user personality labelThe matching relationship between product information purchased with user, to carry out the intelligence of potential customers by artificial intelligence identification modelIdentification helps product dealer to realize precision marketing, avoids being wasted in the time with some invalid clients, improve efficiency, and savesAbout cost, to improve effectiveness of operation.
Specific embodiment
In the following, being described further in conjunction with attached drawing and specific embodiment to the present invention, it should be noted that notUnder the premise of conflicting, new implementation can be formed between various embodiments described below or between each technical characteristic in any combinationExample.
Referring to Fig. 1, it is a kind of flow diagram of potential customers' recognition methods, comprising:
Training sample obtaining step:
S1, product sales data is obtained, the product sales data includes information and the client of the product of client's purchasePersonal information；
S2, the data information for obtaining client from data information platform according to the personal information of client；The data packetsIt includes: interest big data, track big data and/or general lifestyle big data；The interest big data includes clear on the line of userLook at data, the track big data includes the consumption data of user, and the general lifestyle big data includes that the life of user is practisedUsed data；
S3, corresponding label data is established according to the data information of client, the label data includes multiple user personalitiesLabel；
S4, the information for the product that all user personality labels of each client and the client buy is matched and conductOne training sample is stored in training set；
Model construction and training step:
S5, according to user personality label, artificial intelligence identification model is established using machine learning algorithm；
S6, the artificial intelligence identification model is trained using the training sample in training set so that the artificial intelligenceEnergy identification model can judge whether the user is potential customers according to the user personality label inputted；
Identification step:
S7, the data information that user to be identified is obtained from data information platform, build according to the data information of user to be identifiedFound corresponding label data, and by label data input trained artificial intelligence identification model to judge that the user isNo is potential customers.
By the above-mentioned means, first obtaining the passing sale of automobile data of car dealer, sale of automobile data include clientPersonal information and purchase vehicle and brand, then obtained from data information platform according to the personal information of client with useThe relevant data information in family, the data information daily to user are analyzed and processed, and are obtained user personality label and are purchased with userAutomobile information between matching relationship, obtain training data, constructed artificial intelligence known by this part training dataOther model is trained, and artificial intelligence identification model buys specific vehicle by study, the vehicle buyer of particular brand is hadThe some user personality label informations having, thus realize the intelligent recognition that potential customers are carried out by artificial intelligence identification model,Help product dealer realize precision marketing, avoid being wasted in the time with some invalid clients, improve efficiency, save atThis, to improve effectiveness of operation.
It should be noted that interest big data is mainly the browsing data of user, it is available to use according to browsing dataThe hobby at family, its things of interest at this stage etc., and give the corresponding label of user.For example user passes through in TaobaoThe browsing record of search record, number of site in the product type that often browses, search engine, LBS data etc..
Track big data is mainly the consumption data that user really generates, such as product, gap bridge expense are bought in TaobaoRecord etc..Corresponding label is classified and given to the level of consumption of user according to consumption data, or according to consumption numberAccording to determining user's state in which label, such as often, purchase articles for babies is then child-bearing label, buys production relevant to weddingThe consumption data of product is then newly-married label, and having cross-bridge-expense record then is car owner's label etc..
General lifestyle big data is mainly the data of the daily life state of user, according to the daily life state of userData define corresponding label to user, such as program category that user likes watching (military class program-military affairs fan's label,Then the user may be more prone to SUV type automobile), (for example style of wearing the clothes, daily necessity type are partially Japanese for purchase product styleStyle, then the user may be more prone to constitute Nissan Motor brand), purchase the product place of production (such as often purchase importThe product of product or a state, then the user may be more prone to buy the automobile brand of the country), current consumption(user's personality, such as general browsing a few minutes place an order purchase just as resolute type personality or perception spending crowd's label to event；InsteadMultiple browsing places an order as hesitation type or prudent type personality or reasonable consumption crowd label to Bizet) etc..
Particularly, it has then been sticked accordingly to user when data information of the data information platform to user has carried out processingWhen user personality label, the user personality label of user can be directly obtained from data information platform, no longer needs to count itIt is believed that breath is analyzed and processed.For example, Taobao will record the browsing record and shopping record of user, it can be according to the clear of userRecord and shopping record are look to determine that the user likes the product etc. of which class product either what price, that is, is equivalent toCorresponding personality label has been sticked to the user on backstage, has then given user's recommended products according to these personality labels.So,We can obtain personality label's data of user directly from Taobao.
As a preferred embodiment, it is described according to user personality label, it is established using machine learning algorithm artificialIntelligent recognition model specifically:
Based on XGBoost algorithm, using each user personality label as segmentation attribute, several decision trees of building are integrated intoOne artificial intelligent recognition model.
As a preferred embodiment, the pattern function of the artificial intelligence identification model are as follows:
Wherein, t is the quantity of decision tree.
Above-mentioned function needs the new function for guaranteeing to be added, i.e., the new decision tree energy that new user personality label lock is constitutedIt is enough to promote whole expression effect, it promotes expression effect and means that objective function (is exactly to lose plus after new decision treeFunction) value can decline.But since the number of leaf node is more, the risk of over-fitting can be bigger, it is therefore desirable to inhibit leafThe number of child node, so objective function is needed plus a penalty term " omega (ft) ", penalty term is as follows:
Wherein, T is the leaf quantity of current decision tree, and γ is the complexity cost that new leaf node is added and introduces, and w is leafThe weight of child node.
Then artificial complete objective function of intelligent recognition model specifically: (punished by itself loss function and regularizationItem " omega (ft) " is penalized to be added)
The quantity of decision tree is determined by minimum target functional value, seeks local derviation to weight to objective function, and obtaining one canMake the smallest weight of objective function, this weight generation return in objective function, this back substitution result be just to solve for after minimumMinimum target functional value formula described in target function value are as follows:
Wherein, T is the leaf quantity of current decision tree, and γ is the complexity cost that new leaf node is added and introduces.
Obj is represented when we specify the structure of a tree, can at most be reduced in target how much, we can be withIt is called configuration score, this score is the smaller the better, and the smaller structure for representing tree of score is better.
For extending every time, need to enumerate all possible scheme.For some specific segmentation, we will calculate thisThe sum of score and the score of right subtree of the left subtree of a segmentation, then with being compared before division, to calculate this pointThe resulting score of scheme is cut, specific formula is as follows:
When some, which divides resulting score, is less than preset threshold value, we can cut this segmentation to avoid excessively quasi-It closes.In the above manner after the optimum structure to determine artificial intelligence identification model, model is carried out using training sampleTraining can make it that can judge whether user has purchase vehicle demand according to user personality label.
As a preferred embodiment, the information of the product includes the brand and model of product；It is described artificialIntelligent recognition model can judge whether the user is potential customers and the user according to the user personality label inputtedBrang Preference.Namely the artificial intelligent recognition model is made of more categorised decision trees.
As a preferred embodiment, further comprising the steps of: when recognizing user is potential customers, according to the userPersonal information and the data information acquired in the data information platform determine the normal residence of the user for corresponding regionProduct dealer to the user carry out precision marketing.The mode of the normal residence of user can pass through the common shopping online of userShopping area determines under shipping address or line.
As a preferred embodiment, further including negative sample training step:
After training the artificial intelligence identification model, new product sales data is obtained, the product sells numberAccording to the information for the product for including client's purchase and the personal information of client；
The data information of client is obtained from data information platform according to the personal information of client；
Corresponding label data is established according to the data information of client, and label data input is trained artificialRecognition result is obtained in intelligent recognition model；
The recognition result is compared with the effective sale information in the product sales data；
If comparison result is inconsistent, the artificial intelligence identification model is carried out according to the user personality label of the clientAmendment is to improve identification precision.
When obtaining training sample, artificial intelligence can be known using a part of product sales data as Positive training sampleOther model is trained, using another part product sales data as negative training sample to trained artificial intelligence identification modelIt tests, recognition result is compared with effective sale information, when the recognition accuracy of artificial intelligence identification model does not haveWhen reaching preset value, then second training is carried out to it using this portioned product sales data, to improve artificial intelligence identification mouldThe recognition accuracy of type.
As a preferred embodiment, the user personality label includes but is not limited to: interest tags, zone of actionLabel, level of consumption label, division of life span's label, personality label and shopping genre labels.It of courses, in above-mentioned each labelMultiple sub-category tags are further comprised, and above-mentioned label is only relevant example, user personality label can also be according to actualData information marks off other more labels.
After training artificial intelligence identification model, it can periodically send automatically or manually send instruction from data informationPlatform batch obtains the data information of user, is then analyzed and processed to obtain the corresponding user of each user to data informationProperty label, is then identified by artificial intelligence identification model, determines which user is potential customers.Data information platform canTo include but is not limited to Taobao, Jingdone district, search engine, Alipay, Baidu map, Amap and high speed charging system etc.The information platform of some recordable user behaviors.It can be obtained by way of signing cooperation agreement with above-mentioned data information platformThe data information of user, it is ensured that the legitimacy of information.
These data information platforms can reflect the daily life behavior of user, can be determined and be used according to these daily life behaviorsThe personality label at family, such as consumption grade, consumption habit etc..High-end consumption, ordinary consumer are still fallen within for automobile is currentRequire that there is certain economic capability can just go to buy, and often related with life stage, than if any more after childNumber family can consider emphatically purchase vehicle, and therefore, by the analysis of the daily life behavior to user, can estimate that user purchases vehicle canEnergy property helps car dealer to realize precision marketing, avoids time being wasted in one so as to realize the identification of potential customersWith a little invalid clients, improve efficiency, save the cost, to improve effectiveness of operation.
The present invention also provides a kind of electronic equipment, including processor and memory, the memory is stored with executableComputer program, the processor can be read the computer program in the memory and run as described above potential to realizeClient's recognition methods.
In addition, the present invention also provides a kind of computer readable storage medium, the computer-readable recording medium storageThere is executable computer program, the computer program can realize potential customers' recognition methods as described above when running.
The computer-readable recording medium storage has computer program, if method of the invention is with SFU software functional unitForm realize and when sold or used as an independent product, can store in the computer storage medium.Based on suchUnderstand, the present invention realizes all or part of the process in above-described embodiment method, can also instruct phase by computer programThe hardware of pass is completed, and the computer program can be stored in a computer storage medium, which is being locatedIt manages when device executes, it can be achieved that the step of above-mentioned each embodiment of the method.Wherein, the computer program includes computer program generationCode, the computer program code can be source code form, object identification code form, executable file or certain intermediate formsDeng.The computer storage medium may include: any entity or device, record that can carry the computer program codeMedium, USB flash disk, mobile hard disk, magnetic disk, CD, computer storage, read-only memory (ROM, Read-Only Memory), withMachine access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium etc..It should be noted that the content that the computer storage medium includes can be according to legislation and patent practice in jurisdictionIt is required that carrying out increase and decrease appropriate, such as in certain jurisdictions, do not wrapped according to legislation and patent practice, computer storage mediumInclude electric carrier signal and telecommunication signal.
The above embodiment is only the preferred embodiment of the present invention, and the scope of protection of the present invention is not limited thereto,The variation and replacement for any unsubstantiality that those skilled in the art is done on the basis of the present invention belong to institute of the present inventionClaimed range.